Rebase and merge the old SDL2 branch (accelerated scaling)

[PulkoMandy]

Currently GrafX2 uses software scaling to implement the "double", "triple", "wide" and "tall" pixels. This dates back from the SDL1 port (DOS used to tweak videocard parameters directly instead).

This approach works OK for SDL1, but it is CPU hungry. In SDL2 it would make more sense to use the hardware to do it.

This work was started in the original SDL2 branch: ​https://gitlab.com/GrafX2/grafX2/-/tree/sdl2?ref_type=heads but it was not fully complete, and then an independant migration to SDL2 was done. Then an additional support for hardware scaling was added in ​https://gitlab.com/GrafX2/grafX2/-/merge_requests/362 but it is on top of the existing scalers, making things more complex and more confusing.

So the plan is:

• Remove that extra layer of scaling
• Rebase the SDL2 branch or rewrite something along these lines, using hardware scaling to implement double/wide/tall pixels

This has become a bit more complex than it was in DOS, because now we take advantage of the "native" pixels for some things. For example, the layer previews (when hovering the layer bar) and the preview in the file load/save dialog use the full screen resolution